BIS Working Papers No 438 - Monetary and Economic Department
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
BIS Working Papers No 438 Asia’s decoupling: fact, forecast or fiction? by Lillie Lam and James Yetman Monetary and Economic Department December 2013 JEL classification: F62, E32 Keywords: business cycle co-movement, decoupling
BIS Working Papers are written by members of the Monetary and Economic
Department of the Bank for International Settlements, and from time to time by
other economists, and are published by the Bank. The papers are on subjects of
topical interest and are technical in character. The views expressed in them are
those of their authors and not necessarily the views of the BIS.
This publication is available on the BIS website (www.bis.org).
© Bank for International Settlements2013. All rights reserved. Brief excerpts may be
reproduced or translated provided the source is stated.
ISSN 1020-0959 (print)
ISBN 1682-7678 (online)Asia’s decoupling: fact, forecast or fiction?
Lillie Lam and James Yetman1
Abstract
Standard measures of real economic co-movement between Asia-Pacific economies
and those elsewhere had been observed to follow a downward trend, leading some
commentators to suggest that the region was decoupling. However, this process
reversed in response to the International Financial Crisis, and co-movement
increased to historically high levels for some economies. We examine co-movement
patterns and show that these are very sensitive to changes in macroeconomic
volatility over time. Controlling for this, however, co-movement is closely linked to
underlying trade and financial integration. If international links continue to
strengthen in future, co-movement will strengthen in tandem. Decoupling is more a
fiction than a fact or a forecast.
JEL classification: F62, E32
Keywords: business cycle co-movement, decoupling
1. Introduction
In the lead-up to the 2008 International Financial Crisis, many commentators
discussed Asia’s apparent real decoupling from the global economy.2 Despite
growing trade and financial links, the degree of business cycle co-movement
between Asian economies and the major advanced economies appeared to be
decreasing. This view was supported by empirical analysis based on standard
measures of economic co-movement: correlation coefficients and regression
analysis.3
Strong economic co-movement appeared to re-assert itself with a vengeance
once the International Financial Crisis arrived. Using the standard measures that
appeared to suggest decoupling in the lead-up to the crisis, co-movement
1
Bank for International Settlements.
The views expressed here are those of the authors and are not necessarily shared by the Bank for
International Settlements. We thank Mark Spiegel and Philip Turner for helpful comments. Any
remaining errors are solely our responsibility. Corresponding author: Bank for International
th
Settlements, Representative Office for Asia and the Pacific, 78 Floor, Two International Finance
Centre, 8 Finance Street, Central, Hong Kong; email: james.yetman@bis.org; phone:
+852 2878 7152; fax +852 2878 7123.
2
For example, The Economist on May 6, 2008: ”The decoupling debate”; and Bloomberg
Businessweek on March 20, 2008: “Are Asian Economies Decoupling from U.S.?”.
3
Note that we focus on real economic co-movement. We do not explore financial prices, which have
also tended to move more closely together as a result of financial integration. See, for example,
Miyajima et al (2012).
WP438 Asia’s decoupling: fact, forecast or fiction? 1increased strongly once the crisis-period was included in the sample, to historically
high levels for some economies.
In this paper, we re-examine the recent behaviour of standard measures of
business cycle co-movement. First we demonstrate how sensitive the measures are
to the business cycle. During tranquil periods, when there is an absence of
recessions or crises, there is typically little evidence of any co-movement between
Asian economies and the major advanced economies. But during periods of
turbulence, such as recessions or global crises, measures of co-movement are
observed to spike. In this telling, the “great moderation”, a period of low volatility in
business cycle fluctuations prior to the International Financial Crisis, helps to explain
the apparent “decoupling” that commentators observed in the data.
Second, we look for factors that might explain the degree of real co-movement,
focusing on cross-sectional variation across economies. Because co-movement
tends to spike during turbulent periods, we consider volatile and (relatively) tranquil
periods separately. Focusing on three different recent sub-periods (the recession in
2001, the International Financial Crisis, and the period in between) we identify a
number of intuitive factors that explain co-movement. Stronger economic and
financial links, measured in terms of trade flows and cross-holdings of financial
assets, imply stronger economic co-movement. While the size of the estimated
coefficients varies between tranquil and turbulent periods, their signs and statistical
significance do not. Thus the same factors appear to explain co-movement across
the different phases of recent business cycles.
Third, we investigate the prospects for business cycle co-movement. On one
level, the degree of co-movement is dependent on the nature of the global business
cycle. If the global economy continues to recover from the International Financial
Crisis, standard measures of co-movement are likely to remain near current levels.
However, any major deterioration in the global business cycle will likely result in
higher measured co-movement.
We also consider the longer-term implications. Conditioning on the degree of
volatility in the macroeconomy, our results indicate that measures of co-movement
are driven by the strength of underlying economic links. Thus any strengthening of
these links implies greater co-movement in future. The factors that best explain the
degree of business cycle co-movement cross-sectionally have tended to increase
over time. This suggests that, over the longer term, we are likely to see increasing
business cycle co-movement between Asia and the advanced economies, all else
equal. However, this projection depends critically on the continuing increase in
trade and financial links between Asia and the rest of the global economy. In
contrast, if there is a move towards greater economic and financial isolationism, we
should expect to see weakening evidence of co-movement in future.
2. Measuring economic co-movement
There are a number of different measures of economic co-movement. The most
common is the Pearson correlation coefficient, defined as:
2 WP438 Asia’s decoupling: fact, forecast or fiction?1
( yit yi )( y jt y j )
T t
ij ,
1 1
( yit yi ) T 1 t ( y jt y j )
T 1 t
2 2
where yit is the year-over-year percentage change in quarterly real GDP for country
i in period t , yi is the arithmetic mean of yit and j corresponds to some base
economy.4
Another common measure is the coefficient from regression of growth in one
country on growth in the other (see Yeyati and Williams 2012, for example). In the
regression
yit i ij y jt ijt ,
ij is intuitively appealing as it measures the degree to which GDP growth in one
economy is influenced by that in another.
We also consider equivalent measures of co-movement based on the output
gap in countries i and j ,defined by
1
T t
zit z jt
ij
1 1
T 1 t
zit 2
T 1 t
z jt 2
and ij in
zit ij z jt ijt
where zit ln[ xit ] ln[ HP ( xit )] is defined as the output gap, xit is the
seasonally-adjusted level of quarterly real GDP and HP (.) indicates the Hodrick-
Prescott filter with a smoothing parameter of 1600.5,6
4
We consider the US and the G3 economies (weighted average of the euro area, Japan and
the US, based on 2005 GDP and PPP exchange rates) as country j .
5 ij ij
When the Hodrick-Prescott filter is applied across the full sample on which and are
calculated, the mean of the output gap is very close to zero by construction. Hence we do not
include an intercept nor de-trend the output gap when we construct the measures over the full
sample. In other circumstances, for example when we consider rolling samples, we include an
intercept.
6
Alternative approaches to measuring co-movement include measures based on the portion of the
time that economies are simultaneously in the same phase of the business cycle (Berge (2012) and
Harding and Pagan (2002, 2006)) and measures based on the size of spillovers (Diebold and Yilmaz
(2009); applied to Asia in Fujiwara and Takahashi (2012)).
WP438 Asia’s decoupling: fact, forecast or fiction? 3Measures of business cycle co-movement with the US
10-year rolling sample Figure 1
Australia China Hong Kong SAR
ρ
0.6 1.5
β 1.0
θ 0.4
γ 1.0
0.5 0.2
0.5
0.0 0.0
0.0
–0.2
–0.5
–0.4 –0.5
–1.0 –0.6 –1.0
72 77 82 87 92 97 02 07 12 02 04 06 08 10 12 82 87 92 97 02 07 12
India Indonesia Japan
1.0 0.5 1.0
0.5
0.0
0.0 0.5
–0.5
–0.5 0.0
–1.0
–1.0
–0.5
–1.5 –1.5
–2.0 –2.0 –1.0
72 77 82 87 92 97 02 07 12 91 94 97 00 03 06 09 12 67 72 77 82 87 92 97 02 07 12
Korea Malaysia New Zealand
1.0 1.0 1.00
0.75
0.5 0.5
0.50
0.0 0.0
0.25
–0.5 –0.5
0.00
–1.0 –1.0 –0.25
–1.5 –1.5 –0.50
72 77 82 87 92 97 02 07 12 00 02 04 06 08 10 12 91 94 97 00 03 06 09 12
Philippines Singapore Thailand
1.0 2.0 1.0
0.5 1.5 0.5
0.0 1.0 0.0
–0.5 0.5 –0.5
–1.0 0.0 –1.0
–1.5 –0.5 –1.5
85 88 91 94 97 00 03 06 09 12 88 91 94 97 00 03 06 09 12 04 06 08 10 12
Sources: Datastream; national data; authors’ calculations.
Rolling 10-year samples for each of these four standard measures of co-
movement are given in Figure 1 for 12 Asia-Pacific economies against the US
4 WP438 Asia’s decoupling: fact, forecast or fiction?economy.7 This includes ten major Asian economies (China, Hong Kong, India,
Indonesia, Japan, Korea, Malaysia, the Philippines, Singapore and Thailand) as well
as Australia and New Zealand.
With some exceptions, these four measures across the different economies tell
a similar tale of the last 20 years. First, for 10-year rolling samples ending in the
1990s, the measures tend to trend down over time. Second, for samples ending in
the early 2000s, they are generally characterised by low levels of co-movement. A
number of authors in the past picked up on these trends, and interpreted them as
providing evidence of decoupling of Asia from the global economy.8 Following that,
however, measures of business cycle co-movement start to increase, slowly at first,
before jumping to high levels once data for 2008-2009 is included in the rolling
samples.
Note that there is some variation across economies. For example, Australia and
India never saw any large pick-up in co-movement as a result of the crisis. And in
the 1980s Singapore displayed low levels of business cycle co-movement, in
contrast to Australia, Hong Kong, Japan and Korea.
Further, there is some variation across the four measures, especially in the cases
of China and Indonesia, where growth rate-based measures suggest large negative
co-movement during much of the 2000’s. Yet the twelve economies deliver a similar
message: one of apparent decoupling starting in the 1990s following by increasing
co-movement, which a large spike in most measures during the International
Financial Crisis. Comparing the different measures, it is not surprising that they tell
similar stories. In fact, there is a simple relationship between some of the different
measures:
( y jt ) ij ( z jt )
ij ij ; ij
( yit ) ( zit )
where (.) is the standard deviation. Effectively and are different
normalisations of the covariance of growth rates between the two countries, while
and are different normalisations of the covariance of the output gap. For
example, the correlation coefficient ( ) is a measure of the effect of y jt on yit in
terms of standard deviations. It answers the question, conditional on there being a
simple linear relationship between the two growth rates, how many standard
deviations would country i ’s growth rate increase by if there was a one standard
deviation increase in country j ’s growth rate? The regression coefficient ( )
instead focuses on the same question but in terms of units: by how many units
(here defined as percentage points) would country i ’s growth rate increase by if
there was a one unit increase in country j ’s? The ratios of the standard deviations
tend to be relatively stable over time; thus the different measures tell a similar story.
7
The date on the horizontal axis corresponds to the end of the 10-year rolling sample. For equivalent
graphs against a weighted average of the G3 economies (instead of the US), please see Graph A1 in
the appendix.
8
See, for example, Otto et al (2001), Akin and Kose (2008) and Park and Shin (2009).
WP438 Asia’s decoupling: fact, forecast or fiction? 5Measures of business cycle co-movement with the US
Contribution at each quarter Figure 2
Australia China Hong Kong SAR
Measure based on
Growth 20 100
10 Output gap (lhs) 20 4 20
15 75
10 50
5 10 2 10
5 25
0 0
0 0 0 0
–5 –25
–5 –10 –2 –10 –10 –50
82 87 92 97 02 07 12 94 97 00 03 06 09 12 82 87 92 97 02 07 12
India Indonesia Japan
10 20 5 20 20 80
5 10 15 60
0 0
0 0 10 40
–5 –20
–5 –10 5 20
–10 –40
–10 –20 0 0
–15 –30 –15 –60 –5 –20
82 87 92 97 02 07 12 82 87 92 97 02 07 12 82 87 92 97 02 07 12
Korea Malaysia New Zealand
20 80 15 75 15 60
15 60 10 40
10 50
10 40 5 20
5 25
5 20 0 0
0 0
0 0 –5 –20
–5 –20 –5 –25 –10 –40
–10 –40 –10 –50 –15 –60
82 87 92 97 02 07 12 91 94 97 00 03 06 09 12 82 87 92 97 02 07 12
Philippines Singapore Thailand
5 20 15 60
20 100
10 40
0 0
10 50 5 20
–5 –20
0 0
0 0
–10 –40
–5 –20
–15 –60 –10 –50 –10 –40
81 86 91 96 01 06 11 82 87 92 97 02 07 12 97 02 07 12
Shaded areas indicate NBER recession period of the US.
Sources: Datastream; national data; authors’ calculation.
6 WP438 Asia’s decoupling: fact, forecast or fiction?3. Co-movement: the time dimension
To further investigate changes in measures of business cycle co-movement over
time, we focus on the period-by-period contributions to the different measures
without any normalisation. Figure 2 plots ( yit yi )( y jt y j ) and zit z jt at each
point in time for each of the economies in our sample against the US economy.9
These graphs help to explain what we see in the rolling samples. Most of the
time, measures of co-movement are low or close to zero. However, there are
exceptions. During the Asian Financial crisis, the more heavily-affected Asian
economies (Indonesia, Korea, Malaysia and Thailand) grew at a rate below trend
while the US grew at above trend. Hence the contribution to co-movement was
negative. More recently, during the International Financial Crisis, many economies
co-moved very strongly with the US, with the size of the contribution to co-
movement reaching historically high levels in many cases (China, Hong Kong, Japan,
Korea, Malaysia, Singapore and Thailand).10 Given that the standard measures of co-
movement are often based on rolling samples of these variables, as illustrated in
Figure 1, it is not surprising that the presence of the Asian Financial Crisis period in
rolling samples would tend to give the impression of decoupling, while the presence
of the International Financial Crisis in the sample would support the opposite
conclusion.11
Figure 2 also includes shaded bars for periods in which the US economy was in
recession, as defined by the NBER. Here we see a more general pattern. When the
US economy enters a recession, standard measures of co-movement tend to
increase, and the size of the increase appears to be positively related to the severity
of the recession.12 So, many economies saw small increases in co-movement during
the 2001, relatively mild, recession. But in 1990-91 (for Australia, India, Malaysia, the
Philippines and Singapore), and especially in 1981–82 (for Australia, Hong Kong,
Indonesia, Korea and New Zealand), co-movement spiked.
Table 1 supports this interpretation. It displays the median of the average co-
movement measure across the economies in our sample for different sub-samples.
Strikingly, the 2002-2007 period was one of very low co-movement relative to the
recessions that preceded and followed.
9
The sample mean y is based on the full sample of data for country i or j which varies across
economies due to data availability. One difference between the growth rate- and output gap-
based measures is that the former tend to result in a single spike in co-movement at the height of a
period of turbulence while, with the latter measure, this is preceded by another spike. This
difference is partly because the Hodrick-Prescott filter is a two-sided filter: the underlying
observation at time t influences the measure both before and after time t . An equivalent graph
against the G3 economies is given in Graph A2 in the appendix.
10
Antonakakis (2012) also notes that the 2007-2009 US recession was a period of unprecedented
business cycle co-movement.
11
Siklos (2012) argues that the whole notion of decoupling is unhelpful. Instead, it is more
informative to think of variations in the degree of mutual dependence over time.
12
See also Yetman (2010) for related arguments. Similarly Kose et al (2013) report that national
business cycles are more sensitive to developments in the global economy during global recessions
than during global expansions.
WP438 Asia’s decoupling: fact, forecast or fiction? 7Co-movement and the business cycle Table 1
NBER recessions1 Percentage change in Growth-based measure3 Output gap-based measure3
(expansions) real GDP of the US2
1980 Q1–1980 Q3 –2.83 –1.49 0.50
(1980 Q4–1981 Q2) (–2.50)
1981 Q3–1982 Q4 –2.01 4.66 1.13
(1983 Q1–1990 Q2) (1.52) (0.27)
1990 Q3–1991 Q1 –1.02 0.35 0.51
(1991 Q2–2001 Q1) (–1.37) (–0.18)
2001 Q2–2001 Q4 0.73 4.47 0.46
(2002 Q1–2007 Q4) (0.09) (0.24)
2008 Q1–2009 Q2 –4.69 25.22 3.59
1
Recession is defined as a period from the month following the NBER peak to the month of NBER trough. 2 Percentage change of the
3
US real GDP between NBER trough and the preceding NBER peak. For growth-based measure, average of quarter-by-quarter
contribution of the measures within the corresponding recession or expansion; for output gap-based measure, average within one year
before to one year after the corresponding recession or expansion. Median of Australia, China, Hong Kong, India, Indonesia, Japan,
Korea, Malaysia, New Zealand, the Philippines, Singapore and Thailand. Due to missing data, China and Thailand are excluded in
business cycles before 2001; Indonesia before 1982; Malaysia before 1991.
Source: authors’ calculations.
Outside of turbulent periods, Asian economies appear to be largely
“decoupled” and insulated from the advanced economies, based on low levels of
co-movement using standard measures. This is despite the fact that the period we
are examining was one in which Asian economies were becoming increasingly
integrated with the global economy, with trade and financial flows growing, as we
later discuss.
There are a number of possible explanations for the apparent disconnect
between stronger economic links and time-varying co-movement. First, changes in
the degree of co-movement could reflect changes in the strength of underlying
economic links (Kose et al 2003; Furceri and Karras 2008). For example,
strengthening economic links that lead to increases in international risk sharing
would tend to promote greater co-movement. In contrast, stronger trade links may
allow greater specialisation and, depending on the nature of that specialisation
(promoting either intra- or inter- industry trade), may either increase or decrease
co-movement.13 However, it is unlikely that the nature of underlying economic links
can change so quickly as to produce the amount of time-variation in co-movement
that we see in the data.
An alternative explanation is that standard measures of co-movement are
inherently sensitive to the global business cycle. Perhaps the effects of economic
links on co-movement depend critically on the phase of the business cycle or
underlying economic volatility. The so-called “great moderation”, for example, was a
period of relative economic calm by historical standards when Asian economies
found themselves in a relatively benign global economic environment. During such
13
For example, Park (2011) argued that increased intra-regional trade would lead to a self-contained
East Asian region that was relatively sheltered from external shocks. In this sense, continued growth
in intra-regional trade may lead to decoupling.
8 WP438 Asia’s decoupling: fact, forecast or fiction?periods, economies may grow at rates that are relatively close to trend, so that
measures of deviations from trend, such as de-meaned growth rates and output
gaps, tend to be close to zero. In contrast, during periods of turbulence, growth
rates are likely to deviate further from trend, and in the negative direction.14
Additionally, the sign of these deviations is likely to be correlated across countries
due to trade and financial links, so that standard measures of co-movement are
much larger during volatile periods than during other times. We examine these
mechanisms further in the following section.15
4. Co-movement: the cross-section dimension
We next focus on the cross-sectional dimension, and examine what might explain
the relative degree of co-movement for the economies in Asia-Pacific. Given the
foregoing analysis, outlining how much co-movement varies between turbulent
periods and relatively tranquil periods, we divide recent years into three different
periods: the 2001 recession, the International Financial Crisis and the period in
between. We take these recent periods as being the most informative for the
prospects of co-movement in future, given that economic links between economies
tend to evolve slowly over time. Each period is examined separately, using ordinary
least squares. As a robustness check, we also estimate across the three periods
together using seemingly unrelated regression methods, and obtain very similar
results to those reported below.16
4.1 International Financial Crisis of 2008–2009
Starting with the most recent sub-period, for the International Financial Crisis
episode we compute the average of the quarter-by-quarter contribution to our four
measures of business cycle co-movement outlined in section 2. For and we
focus on the period 2008 Q1–2009 Q2, while for and we focus on the longer
period of 2007 Q1–2010 Q2. For the first two measures, our period corresponds to
the recession as defined by NBER business cycle dates.17 For the latter two
measures, the longer examination period reflects the fact that measures based on a
two-sided filter, like the Hodrick-Prescott filter used here, imply that the underlying
14
For example, Leduc and Liu (2013) argue that increased uncertainty may manifest itself as a
downward demand shock. Their estimates suggest that uncertainty shocks during the international
financial crisis accounted for at least a one percentage point increase in unemployment in the US,
15
Yetman (2011) suggests a measure of business cycle co-movement that is relatively insensitive to
the amount of macroeconomic volatility given by t ( yit yi ) ( y jt y j ) . See, also, Wälti
ij
(2009) for a similar approach.
16
The results from the seemingly unrelated regression are contained in Table A2 in the appendix.
Seemingly unrelated regression allows estimates of coefficients to vary across equations but offers
efficiency gains from modelling the correlation in errors across the three samples.
17
In accordance with NBER business cycle reference dates, the recession period in this paper is
defined as starting with the quarter in which the month following the NBER peak falls, and ending
with the quarter in which the NBER trough falls. For example, December 2007 is the month of the
latest peak and June 2009 is the latest trough. So the recession period is defined as the period of
January 2008–June 2009 (i.e. 2008 Q1–2009 Q2).
WP438 Asia’s decoupling: fact, forecast or fiction? 9observation at time t influences the measure both before and after time t . Our
approach yields four measures across 12 economies, for a total of 48 observations.
We examine 24 possible variables that might explain the degree to which Asian
economies co-move with the US, listed in table 2. Except where otherwise specified,
we measure these variables as of the end of 2007, before the greatest effects of the
crisis were felt in the Asian region.
Variables to explain co-movement during the International Financial Crisis Table 2
Units Observations Mean Standard deviation
1. Trade openness (end-2007)
Trade openness = exports + imports % of GDP 10 114.18 114.39
Manufacturing exports % of exports 12 57.16 23.02
2. Financial openness (end-2007)
Current account % of GDP 12 5.82 9.63
Net foreign assets % of GDP 12 13.85 96.48
Gross foreign assets % of GDP 12 216.56 379.17
Gross foreign liabilities % of GDP 12 202.71 295.21
Foreign holdings of US assets % of GDP 12 25.34 30.46
Foreign holdings of US LT debt % of GDP 12 15.44 14.98
Foreign holdings of US ST debt % of GDP 12 1.32 0.98
Foreign holdings of US equity % of GDP 12 8.58 18.27
US holdings of foreign assets % of foreign GDP 12 16.80 16.48
US holdings of foreign LT debt % of foreign GDP 12 2.31 2.44
US holdings of foreign ST debt % of foreign GDP 12 0.30 0.54
US holdings of foreign equity % of foreign GDP 12 14.19 16.01
Foreign banks’ share of US credit % of total claims 5 2.78 5.16
US banks’ share of foreign credit % of total claims 12 13.08 6.62
Private sector credit % of GDP 12 90.31 39.35
3. Monetary and fiscal policy framework (end-2007)
Exchange rate peg = 1 dummy variable 12 0.25 0.45
Foreign exchange reserves % of GDP 12 35.12 28.14
Exchange rate volatility (2001–07) % 12 6.15 3.99
Inflation target = 1 dummy variable 12 0.50 0.52
Average inflation (2001–07) % 12 2.91 2.66
Inflation volatility (2001–07) % 12 5.08 3.39
Government revenue % of GDP 12 24.56 6.09
Government spending % of GDP 12 23.25 6.94
Government debt % of GDP 12 49.17 50.01
Sources: Lane and Milesi-Ferretti (2007); IMF IFS; IMF WEO; US Treasury International Capital data; national data; BIS consolidated banking
statistics.
Included in the set of explanatory variables are two that focus on trade
openness: total trade as a percentage of GDP, and manufacturing exports as a share
of total exports. Next are fifteen measures associated with financial openness. We
consider the current account, which is a measure of net financing needs of the
economy. Next is net foreign assets, gross foreign assets and gross foreign liabilities
10 WP438 Asia’s decoupling: fact, forecast or fiction?as measures of foreign exposures. Given that the US was a focal point of the crisis,
we measure exposures with the US economy based on Treasury International
Capital System (TIC) data: total foreign holdings of US assets and total US holdings
of foreign assets, each as a share of GDP, as well as break-downs of these variables
into long-term debt, short-term debt and equity. We also consider cross-border
banking exposures based on BIS consolidated banking statistics18 and private sector
credit as a percentage of GDP to measure of the size of domestic financial markets.
Finally, we include measures that are symptomatic of the policy framework. In
terms of monetary policy, dummy variables are used to capture whether the
economy had a pegged exchange rate, or a formal inflation target. We also examine
the size of foreign exchange reserves, daily exchange rate volatility against the US
dollar over 2001-2007 and the level and volatility of inflation over 2001-2007. For
fiscal policy, we examine government revenue, spending and debt, each as a
percentage of GDP.
These measures are intended to include a number of domestic, as well as
international, factors that have been used elsewhere to try to explain the degree to
which different economies were affected by the crisis, for example in Cechetti, King
and Yetman (2011).19
In order to assess the importance of each of these variables in explaining the
degree of co-movement, we consider each in turn in a regression with an intercept
in our panel of 12 economies and 4 different measures of co-movement during the
crisis. Fixed effects are included for each of the different measures. All explanatory
variables are normalised by their standard deviation so that the magnitude of the
coefficient may be compared across variables: it is a measure of the effect of a one-
standard deviation increase in the explanatory variable on the measures of business
cycle co-movement. Results are given in Table 3.
The results reveal a remarkable number of highly statistically significant
variables. In terms of trade openness, both variables are highly significant. Stronger
trade links implied stronger co-movement during the crisis. For financial openness,
all variables except US holdings of foreign debt and US banks’ share of foreign
credit are positive and highly significant. Domestic holdings of US assets of any kind
appear to have been a powerful source of contagion during the crisis. In contrast,
only US holdings of domestic equity had a large effect. Additionally, high levels of
domestic credit appear to have been correlated with higher levels of contagion,
perhaps reflecting heightened vulnerability of the domestic economy to foreign
shocks.
18
Foreign banks’ share of US credit is defined as the share of a given country’s consolidated foreign
claims vis-à-vis the US (on immediate borrow basis by nationality) to all reporting countries claims
vis-à-vis the US. This variable is available for Australia, Hong Kong, India, Japan and Singapore, all
of whom are BIS reporting economies. The US banks’ share of foreign credit is defined as the share
of consolidated foreign claims on immediate borrower basis in a given country by US
headquartered banks of all reporting banks. Since US is one of the BIS reporting economies, this
variable is available for all economies in our sample.
19
Cecchetti et al (2011) ask a different, albeit related, question to the one examined here. First they
construct a measure of how well different economies performed during the crisis, conditional on
the historical dependence between those countries and the US. As a second step, they then
consider similar variables to the ones we use here to see which factors might explain the relative
performance of different economies, conditional on their historical dependence. Here we are not so
concerned about whether economies did better or worse than might be expected but rather, in
absolute terms, how well they performed.
WP438 Asia’s decoupling: fact, forecast or fiction? 11Explaining co-movement during the International Financial Crisis Table 3
Coefficient P-value Observations R2
1. Trade openness (end-2007)
Trade openness = exports + imports 0.67 0.007 40 0.45
Manufacturing exports 0.75 0.002 48 0.46
2. Financial openness (end-2007)
Current account 0.85 0.000 48 0.49
Net foreign assets 0.98 0.000 48 0.54
Gross foreign assets 0.75 0.002 48 0.46
Gross foreign liabilities 0.65 0.005 48 0.44
Foreign holdings of US assets 0.96 0.000 48 0.53
Foreign holdings of US LT debt 0.99 0.000 48 0.54
Foreign holdings of US ST debt 0.77 0.001 48 0.47
Foreign holdings of US equity 0.76 0.002 48 0.46
US holdings of foreign assets 0.80 0.001 48 0.48
US holdings of foreign LT debt 0.05 0.859 48 0.36
US holdings of foreign ST debt -0.24 0.200 48 0.37
US holdings of foreign equity 0.83 0.001 48 0.48
Foreign banks’ share of US credit 0.82 0.043 20 0.46
US banks’ share of foreign credit -0.13 0.645 48 0.36
Private sector credit 0.78 0.002 48 0.47
3. Monetary and fiscal policy framework (end-2007)
Exchange rate peg = 1 0.53 0.062 48 0.41
Foreign exchange reserves 0.96 0.000 48 0.53
Exchange rate volatility (2001-07) -0.43 0.082 48 0.39
Inflation target = 1 -0.61 0.030 48 0.43
Average inflation (2001–07) -1.17 0.000 48 0.61
Inflation volatility (2001–07) -0.70 0.005 48 0.45
Government revenue 0.41 0.153 48 0.39
Government spending -0.15 0.600 48 0.36
Government debt 0.57 0.017 48 0.42
Source: authors’ calculations.
Regarding the policy framework, the general picture is that measures
associated with exchange rate flexibility (not having a currency peg, having a low
level of foreign exchange reserves, allowing high exchange rate volatility) appeared
to lower the level of co-movement during the crisis, implying that the ability of the
exchange rate to adjust and to act as a shock absorber was important in sheltering
domestic economies from external shocks and lowering co-movement during the
crisis. In contrast, having an inflation target was associated with weaker co-
movement. Our results suggest that not all nominal anchors are equal, at least in
the context of crisis periods. However, we find that lower average inflation or
inflation volatility are associated with higher co-movement. This may be because
the economies with the lowest and most stable inflation (Australia, Hong Kong,
Japan, New Zealand and Singapore) are also those with the strongest international
financial links. For fiscal policy, the size of the government (in terms of revenue or
12 WP438 Asia’s decoupling: fact, forecast or fiction?spending as a percentage of GDP) offers little explanatory power. In contrast, the
size of government debt matters. Perhaps an economy saddled with high debt
levels encounters more difficulties in implementing counter-cyclical fiscal policy to
dampen the business cycle during turbulent periods.
Overall, measures of openness had very predictable effects on business cycle
co-movement during the crisis: higher levels of openness resulted in higher levels of
co-movement. Additionally, exchange rate rigidity and high levels of fiscal debt
limited the policy options to absorb shocks, increasing measured co-movement.20
We also considered some robustness checks. Given that the four sets of
estimates of { , , , } that are used to construct our dependent variable are
likely to be correlated, we split our sample in six different ways, by taking every
possible pair of the four measures. The results (available on request) are surprisingly
robust. First, in no single case did a variable that was reported as being statistically
significant in Table 3 change sign. In just one case of each of US holdings of foreign
long term debt, US banks’ share of foreign credit and government spending (all
insignificant in Table 3) the sign of the estimated coefficient changed, and each of
these is highly insignificant (p-values exceeding 0.80). Second, in only a few cases
are statistically significant results in Table 3 no longer significant in one or more of
the split samples. For example, if we define significance at the 10% level, we lack
significance in the split samples for foreign banks’ share of US credit, the exchange
rate peg dummy and inflation targeting dummy (for each variable insignificant in
three of six cases), exchange rate volatility (insignificant in five cases), inflation
volatility and government debt (each insignificant in one case). Third, overall, of the
statistically significant variables at the 10% level in Table 3, 87% of all the results on
split samples have the same sign and are also statistically significant at the 10%
level.21 The remaining 13% have the same sign but are statistically insignificant.22
20
We also considered combinations of these factors, to see if we could find a small number of
variables that together give a parsimonious explanation of co-movement. However, there are no
robust set of regressors: testing up or testing down, first eliminating relatively weak regressors from
the sample or jointly choosing sets two or three regressors based on goodness-of-fit criteria all
offer different combinations from among the significant variables identified above as to which
smaller set of regressors is a good predictor of business cycle co-movement during the crisis.
21
At the 5% level, this percentage is also 87%; of the remaining cases, 9% are statistically significant at
the 10% level while 4% are insignificant.
22
We also considered various pairs and triplets of variables that are likely to be closely related,
although these results were more mixed. If we include both trade variables together, for example,
trade openness becomes insignificant (p-value of 0.16) while manufacturing exports remain highly
significant (p-value of 0.01). In terms of financial openness, both gross foreign assets and gross
foreign liabilities are highly significant when examined together. However, gross foreign liabilities
ceases to be statistically significant when paired with private sector credit. Likewise, while both US
holdings of foreign assets and foreign holdings of US assets are significant when considered on
their own, only foreign holdings of US assets are significant when considered together. Considering
the policy framework variables, when examined jointly, both foreign exchange reserves and having
an exchange rate peg are statistically significant. However, conditioning on the level of foreign
exchange reserves, a pegged exchange rate decreases the degree of co-movement.
WP438 Asia’s decoupling: fact, forecast or fiction? 13Variables to explain co-movement during the 2001 recession Table 4
Units Observations Mean Standard deviation
1. Trade openness (end-2000)
Trade openness = exports + imports % of GDP 9 112.75 85.63
Manufacturing exports % of exports 11 63.46 22.93
2. Financial openness (end-2000)
Current account % of GDP 12 2.61 5.01
Net foreign assets % of GDP 12 -5.12 72.80
Gross foreign assets % of GDP 12 133.98 210.58
Gross foreign liabilities % of GDP 12 139.10 151.01
Foreign holdings of US assets % of GDP 12 16.50 24.98
Foreign holdings of US LT debt % of GDP 12 11.10 14.41
Foreign holdings of US equity % of GDP 12 5.40 11.22
Foreign banks’ share of US credit % of total claims 3 7.08 11.54
US banks’ share of foreign credit % of total claims 12 13.65 6.39
Private sector credit % of GDP 10 104.97 51.43
3. Monetary and fiscal policy framework (end-2000)
Exchange rate peg = 1 dummy variable 12 0.33 0.49
Foreign exchange reserves % of GDP 12 25.07 25.03
Exchange rate volatility (1991-2000) % 12 12.35 7.61
Inflation target = 1 dummy variable 12 0.42 0.51
Average inflation (1991–2000) % 12 5.27 3.84
Inflation volatility (1991–2000) % 12 14.39 17.43
Government revenue % of GDP 12 22.33 7.34
Government spending % of GDP 12 23.63 7.15
Government debt % of GDP 11 56.99 38.57
Sources: Lane and Milesi-Ferretti (2007); IMF IFS; IMF WEO; US Treasury International Capital data; national data; BIS consolidated
banking statistics.
4.2 2001 recession
We now consider the previous recessionary period. Similar to the analysis for the
International Financial Crisis, we compute the average quarter-by-quarter
contribution over the period of 2001 Q2–2001 Q4 to and and over the period
of 2000 Q2–2002 Q4 to and . The set of explanatory variables, listed in Table 4,
is as of the end of 2000 and includes the same variables as those in Table 2 except
where data are unavailable.
The results are given in Table 5. Remarkably, even though the 2001 period was
much less volatile than the more recent crisis, the same variables are in general
statistically significant – even if the coefficients are smaller, by a factor of about 2.
That is, more open economies – measured in terms of trade flows or dependence
14 WP438 Asia’s decoupling: fact, forecast or fiction?on manufacturing exports – tended to co-move more strongly. Greater financial
openness, in terms of gross or net positions, or holdings of US assets, was also
correlated with greater co-movement.23 And factors associated with policy regimes
that allow for exchange rate flexibility – such as inflation targeting and volatile
exchange rates – are associated with lower levels of co-movement. Further, variable
values consistent with economies having the scope to easily respond to external
shocks via policy were correlated with less co-movement.
Explaining co-movement during the 2001 recession Table 5
Coefficient P-value Observations R2
1. Trade openness (end-2000)
Trade openness = exports + imports 0.36 0.001 36 0.54
Manufacturing exports 0.28 0.000 44 0.44
2. Financial openness (end-2000)
Current account 0.48 0.000 48 0.53
Net foreign assets 0.52 0.000 48 0.58
Gross foreign assets 0.45 0.001 48 0.49
Gross foreign liabilities 0.37 0.002 48 0.40
Foreign holdings of US assets 0.51 0.000 48 0.57
Foreign holdings of US LT debt 0.51 0.000 48 0.56
Foreign holdings of US equity 0.49 0.001 48 0.54
Foreign banks’ share of US credit -0.52 0.023 12 0.82
US banks’ share of foreign credit -0.05 0.545 48 0.23
Private sector credit 0.26 0.000 40 0.45
3. Monetary and fiscal policy framework (end-2000)
Exchange rate peg = 1 0.49 0.000 48 0.54
Foreign exchange reserves 0.55 0.000 48 0.62
Exchange rate volatility (1991-2000) -0.18 0.059 48 0.27
Inflation target = 1 -0.37 0.000 48 0.40
Average inflation (1991–2000) -0.20 0.045 48 0.28
Inflation volatility (1991–2000) -0.13 0.073 48 0.25
Government revenue 0.03 0.829 48 0.23
Government spending -0.14 0.133 48 0.26
Government debt 0.14 0.107 44 0.23
Source: authors’ calculations.
One common argument about the international financial crisis is that a collapse
in trade volumes played a critical role in the propagation of the crisis.24 Comparing
our results over these two sub-samples suggests that, while trade flows were a
vector of contagion, their role during the recent crisis was nothing extraordinary.
23
The only exception to this is that foreign banks’ share of US credit was significantly negatively
correlated with co-movement during this period. Note, however, that this result is based on data
for only 3 countries due to data availability.
24
See, for example, Chor and Manova (2012).
WP438 Asia’s decoupling: fact, forecast or fiction? 15Most factors that help to explain co-movement have larger coefficients in the recent
crisis relative to the earlier recession but, proportionately, there is no outsized
increase in the role of trade openness.
Overall, the results across the two episodes suggest that variables reflecting
strong economic links provide information that is highly predictive of the nature of
co-movement during volatile periods.
4.3 Great Moderation of 2002–2007
Finally, we repeat the analysis focusing on a relatively tranquil period. We again
construct the average quarter-by-quarter contribution over the period of 2002 Q1–
2007 Q4 to and and over the period of 2003 Q1–2006 Q4 to and .25 The
set of explanatory variables is the same as those used for the 2001 recessions, i.e. as
of the end of 2000, listed in Table 4.26
The results are shown in Table 6. Curiously, the same variables that were
statistically significant during the more volatile periods are also significant during
this relatively tranquil period, have the same sign and have very similar p-values.
One important difference, however, is that the size of the coefficients is much
smaller during this less volatile period, by a factor of around 5.27
By dividing up the 2001-2009 period into three sub-periods, and focusing on
each of these individually, we have shown that measures of co-movement that are
associated with economic links in terms of trade integration, financial openness and
monetary and fiscal policy flexibility work well to explain co-movement during all
three sub-periods. However, the magnitude of the coefficients varies widely
between the different episodes, with more volatile periods being associated with
larger coefficients. If instead we were just to focus on all the periods together, we
would be combining episodes during which the relationships between economic
links and co-movement vary widely, even as their economic and statistical
significance remains strong.
25
In contrast to the previous two sub-samples, in this case our output gap-based measures are
examined over a shorter period than the growth-based measures, to exclude the effects of volatile
data at either end.
26
The same set of explanatory variables, but as of end-2002, was also considered, yielding very similar
results; see Table A1 in the appendix.
27
Pula and Peltonen (2011) argued that trade data overstate trade openness and analysed decoupling
using an international input-output table which focused on bilateral trade and production linkages.
Their results, based on data up to 2006, argued against decoupling but suggested that emerging
Asia is less “coupled” with the advanced economies than trade data would imply.
16 WP438 Asia’s decoupling: fact, forecast or fiction?Explaining co-movement during the Great Moderation of 2002–07 Table 6
Coefficient P-value Observations R2
1. Trade openness (end-2000)
Trade openness = exports + imports 0.09 0.004 36 0.48
Manufacturing exports 0.05 0.003 44 0.24
2. Financial openness (end-2000)
Current account 0.07 0.000 48 0.32
Net foreign assets 0.12 0.000 48 0.66
Gross foreign assets 0.12 0.000 48 0.65
Gross foreign liabilities 0.11 0.000 48 0.56
Foreign holdings of US assets 0.11 0.000 48 0.54
Foreign holdings of US LT debt 0.11 0.000 48 0.58
Foreign holdings of US equity 0.09 0.000 48 0.45
Foreign banks’ share of US credit -0.11 0.046 12 0.81
US banks’ share of foreign credit 0.00 0.914 48 0.14
Private sector credit 0.07 0.001 40 0.30
3. Monetary and fiscal policy framework (end-2000)
Exchange rate peg = 1 0.08 0.007 48 0.34
Foreign exchange reserves 0.11 0.000 48 0.60
Exchange rate volatility (1991–2000) -0.08 0.002 48 0.22
Inflation target = 1 -0.15 0.000 48 0.34
Average inflation (1991–2000) -0.04 0.014 48 0.19
Inflation volatility (1991–2000) -0.03 0.023 48 0.17
Government revenue 0.01 0.762 48 0.14
Government spending -0.01 0.503 48 0.14
Government debt 0.06 0.000 44 0.30
Source: authors’ calculations.
5. The future of co-movement
The foregoing analysis has illustrated how macroeconomic co-movement, as
commonly measured, varies with the degree of macroeconomic volatility. During
turbulent periods, co-movement is strong compared with relatively tranquil periods.
However, regardless of the degree of volatility, underlying economic links help to
explain the degree of co-movement across our panel of economies.
Looking forward, evidence of either decoupling or higher levels of co-
movement is likely to reflect the global business cycle. If global growth is relatively
stable, and major macroeconomic crises are avoided, then standard measures are
WP438 Asia’s decoupling: fact, forecast or fiction? 17likely to indicate low levels of co-movement. In contrast, periods of turbulence are
likely to result in higher levels of co-movement.28
Cross-sectionally, however, the strength of underlying economic links is highly
correlated with the degree of co-movement, during both volatile and tranquil
periods. Conditioning on the level of macroeconomic volatility, then, the direction
of future co-movement is therefore likely to reflect the strength of underlying
economic links between economies.
The message from data on economic links is generally consistent with
continued strong co-movement. Focusing on trade openness, exports as a
percentage of GDP (Figure 3, top row) have tended to strengthen or remain flat for
most economies in the region. In contrast, the importance of manufacturing for
exports appears to be declining in most countries, although it remained above 60%
of exports for China, Hong Kong, Japan, Korea, the Philippines and Thailand as of
2010. However, current accounts generally remain positive and large for many
regional economies, although these may be expected to fall over the long run.
In terms of other measures of financial openness, gross international exposures
have tended to trend up over time. While these shrank somewhat during the
International Financial Crisis, this has since been reversed. Gross international
exposures are at or are close to all-time highs as a percentage of GDP in the latest
available data for all economies in our sample except Indonesia and the Philippines,
two countries whose exposures never fully recovered from the Asian Financial Crisis
in the 1990s (Figure 3, middle row). Gross and net positions based on TIC data tell a
more nuanced story; some categories have seen strong growth, and others are
declining. But overall, financial links between Asian economies and the US remain
strong.29 Further, many measures of domestic credit are currently at high levels by
historical standards, exceeding 100% of GDP in Australia, China, Hong Kong, Japan,
Korea, Malaysia, New Zealand and Singapore, suggesting that regional economies
remain vulnerable to external shocks.
In terms of the scope for domestic policy responses in the face of external
shocks, there is greater room for optimism. While foreign exchange reserves have
trended up in recent years (Figure 3, bottom row), as policymakers in the region
tended to dampen exchange rate appreciation pressures, this trend is unlikely to
continue indefinitely.30 Any change away from foreign exchange intervention is
likely to be one towards greater exchange rate volatility, which would be associated
with less co-movement. Finally, net government debt remains small and
manageable for most economies in the Asia-Pacific region. With the notable
exception of Japan, net debt exceeds 50% of GDP only in India (where it is trending
down) and Malaysia (52% in 2011). Thus there might be scope for a strong fiscal
28
Leduc and Spiegel (2013) argue that the decline in co-movement in the aftermath of the recent
crisis has been large by historical standards and suggest that this can be explained by monetary
policy in some economies being constrained by the zero lower bound. This effect might be
expected to reverse when monetary policy normalises.
29
See, also, Elekdag et al (2012) for a discussion of growing financial linkages between Asia and the
advanced economies. Financial integration leads to stronger co-movement of asset prices, as
Miyajima et al (2012) discuss. This may be one of the channels by which stronger financial links lead
to stronger business co-movement, although evaluation of the precise channels driving co-
movement is beyond the scope of the current study.
30
See Filardo and Yetman (2012) for a discussion of the challenges associated with the continued
accumulation of foreign exchange reserves in Asia.
18 WP438 Asia’s decoupling: fact, forecast or fiction?response to external shocks to the business cycle, as was seen during the
International Finance Crisis, in contrast to many other regions of the world.
Trade and financial exposures of Asia-Pacific economies
As a percentage of GDP Figure 3
Total trade1
100
400
80
300
60
200
40
100
20
0 0
1991 1994 1997 2000 2003 2006 2009 2012 1991 1994 1997 2000 2003 2006 2009 2012
AU NZ JP KR HK ID MY PH TH
Gross foreign exposure2
250
2,000 200
200
1,500 150
150
1,000 100
100
500 50
50
0 0 0
1971 1976 1981 1986 1991 1996 2001 2006 2011 1993 1996 1999 2002 2005 2008 2011
Lhs Rhs CN ID IN KR PH TH
AU JP MY NZ HK SG
Foreign exchange reserves3
100 25
80 20
60 15
40 10
20 5
0 0
1991 1994 1997 2000 2003 2006 2009 2012 1991 1994 1997 2000 2003 2006 2009 2012
CN HK KR MY SG TH AU ID IN JP NZ PH
AU = Australia; CN = China; HK = Hong Kong SAR; ID = Indonesia; IN = India; JP = Japan; KR = Korea; MY = Malaysia; NZ = New Zealand;
PH = Philippines; SG = Singapore; TH = Thailand.
1
Sum of imports and exports. 2 Sum of foreign assets and liabilities. 3
As of 2011 Q4 for Singapore; 2012 Q1 for China and India;
2012 Q3 for Indonesia and Japan; and 2012 Q2 for others.
Sources: Lane and Milesi-Ferretti (2007); IMF IFS; World Bank; national data.
WP438 Asia’s decoupling: fact, forecast or fiction? 19Overall, the evidence points to continued strong links between the economies
of Asia-Pacific and the advanced economies. Thus it is highly unlikely that we will
see the Asia-Pacific region decoupling from developments elsewhere in the
foreseeable future. Conditional on underlying macroeconomic volatility, advanced
economies outside the region are likely to continue to have large effects on the
economies in Asia-Pacific.
6. Conclusions
Is Asia’s decoupling a fact, a forecast or a fiction? The evidence that we have
presented here suggests that it is closest to being a fiction. First, evidence in the
past of decoupling was heavily skewed by macroeconomic volatility. We have
shown that standard measures imply that co-movement is strongest during
turbulent periods, and indicate little co-movement during relatively tranquil periods.
This dynamic explains most of the past evidence of apparent decoupling. Second,
we show that cross-sectional variation in the degree of co-movement can be
explained by underlying economic links among economies in terms of trade and
financial flows, as well as the scope for domestic policymakers to respond to
external shocks. These relationships are statistically significant, both in turbulent
times and tranquil times, and imply that any long-term forecast of decoupling
requires matching forecasts of decreasing trade and financial links, and/or increased
policy independence in future. While such outcomes, consistent with a change from
internationalisation to isolationism, are possible, they imply a reversal of current
trends that seems unlikely. Thus Asian economies are liable to continue to co-move
closely with the world’s major economies in future.
However, standard measures of co-movement may continue to mislead,
indicating decoupling where none is present. For example, these measures are often
reported based on rolling samples. Our results suggest that, if the global economy
continues to recover from the International Financial Crisis, standard measures of
co-movement are likely to remain near current levels in the near term. But there will
be a discrete drop in measured co-movement in future, when the exceptionally
turbulent period of the International Financial Crisis drops out of the sample period.
Past experience suggests that this will be mis-interpreted as evidence of decoupling,
even if underlying economic links between Asia and the rest of the global economy
continue to strengthen.
20 WP438 Asia’s decoupling: fact, forecast or fiction?Appendix
Measures of business cycle co-movement with G31
10-year rolling sample Figure A1
Australia China Hong Kong SAR
ρ
0.6
β 1.5
θ 1
0.4
γ
1.0 0.2
0
0.5 0.0
–0.2
–1
0.0
–0.4
–0.5 –0.6 –2
77 82 87 92 97 02 07 12 02 04 06 08 10 12 82 87 92 97 02 07 12
India Indonesia Japan
1.5 0.5 1.0
1.0
0.0
0.5 0.5
–0.5
0.0 0.0
–1.0
–0.5
–0.5
–1.0 –1.5
–1.5 –2.0 –1.0
77 82 87 92 97 02 07 12 91 94 97 00 03 06 09 12 77 82 87 92 97 02 07 12
Korea Malaysia New Zealand
1.5 1.5
2
1.0
1.0
0.5 1
0.5
0.0
0
0.0
–0.5
–1.0 –1 –0.5
77 82 87 92 97 02 07 12 00 02 04 06 08 10 12 91 94 97 00 03 06 09 12
Philippines Singapore Thailand
2 1.0
2
1 0.5
1
0 0.0
0
–1 –0.5
–2 –1 –1.0
85 88 91 94 97 00 03 06 09 12 88 91 94 97 00 03 06 09 12 04 06 08 10 12
1
Weighted average of Euro area, Japan and the United States based on 2005 GDP and PPP exchange rate.
Sources: Datastream; national data; authors’ calculations.
WP438 Asia’s decoupling: fact, forecast or fiction? 21Measures of business cycle co-movement with G31
Contribution at each quarter Figure A2
Australia China Hong Kong SAR
Measure based on
yoyAU 20 6 30
10 gapAU 20 100
15
4 20
5 10 10 50
2 10
5
0 0 0
0 0
0
–5 –5 –2 –10 –10 –50
82 87 92 97 02 07 12 94 97 00 03 06 09 12 82 87 92 97 02 07 12
India Indonesia Japan
25 100
10 10
5 10
20 80
5 5 15 60
0 0
0 0 10 40
5 20
–5 –10
–5 –5
0 0
–10 –20 –10 –10 –5 –20
82 87 92 97 02 07 12 82 87 92 97 02 07 12 82 87 92 97 02 07 12
Korea Malaysia New Zealand
20 80 20 80 15 60
10 40
15 60 15 60
5 20
10 40 10 40
0 0
5 20 5 20
–5 –20
0 0 0 0 –10 –40
–5 –20 –5 –20 –15 –60
82 87 92 97 02 07 12 91 94 97 00 03 06 09 12 82 87 92 97 02 07 12
Philippines Singapore Thailand
10 20 20 80
20 100
5 10 15 60
0 0 10 50 10 40
–5 –10 5 20
0 0
–10 –20 0 0
–15 –30 –10 –50 –5 –20
82 87 92 97 02 07 12 82 87 92 97 02 07 12 97 02 07 12
Shaded areas indicate NBER recession period of the US.
1
Weighted average of Euro area, Japan and the United States based on 2005 GDP and PPP exchange rate.
Sources: Datastream; national data; authors’ calculations.
22 WP438 Asia’s decoupling: fact, forecast or fiction?You can also read
